Avoiding Equipment Noise – AES 2010
November 4, 2010, Audio Engineering Society Convention—San Francisco—Bill Whitlock, president of Jensen Transformers presented “Is your equipment design a noise problem waiting to happen?” at the AES conference. In general, problems with audio equipment is experienced as noise, and is very easily noticed. Taking care of the interfaces and the grounding tends to solve the noise problems.
All of the issues related to noisy equipment can be traced to some failure to address physics. Analog signals need a dynamic range of about 120 dB so noise prevention starts with the interfaces to prevent noise coupling to other equipment. Ground noise—hum, buzz, pops, clicks, and other audio discontinuities—are not due to the grounding, but to other defects in the design.
The grounding in equipment is designed for safety and to meet various code requirements. The ground is supposed to be a safety fail-safe, and will perform its function when the neutral and safety are joined together at the service entrance. Defects to the safety include insulation failure and subsequent shorts or opens due to open wires or someone disconnecting the ground plug.
When a ground fault occurs, it conducts across all of the connected equipment. This fault current then trips the breaker or ground fault interrupter as a function of current and time. the higher the current, the shorter the trip time. One item to consider is that the safety ground is not the same as earth ground. Earth is a connection for lightening protection and presents the potential for other fault current paths to ground. Earth ground is not equal to a zero volt potential.
Some users have decided that the AC wiring is the main cause for noise in audio equipment. To eliminate this alleged noise source, they have eliminated the ground connections in the power plugs. This action creates a risk for shocks when users are working across different feeder circuits. Even with the ground lines intact, there will always be a voltage differential between two feeds because the wiring itself acts like a one turn transformer.
To reduce the coupling to ground requires the design include a requirement that the line, neutral, and ground wires be twisted throughout all their runs. Untwisted lines are more susceptible to high frequency harmonics, those signals with a high di/dt, such as those from a light dimmer. Unfortunately, you cannot “short out” this noise, so it is better to find the source and reduce the noise at the source.
After you have created a clean power source, it’s time to look at the interfaces. Ideally the interface is differential. A balanced interface does not need a ground, so ground noise is automatically eliminated. A balanced impedance is parallel to the ground and is defined as equal impedances to ground. On an unbalanced line, one side is grounded.
Impedances don’t have to be matched. The idea for matching came from the telephone companies who had a passive system that needed maximum power transfer. The classic 600 ? matched impedances are designed for a voltage driven system that is terminated at the ends of a 4,000 ft. transmission line. This low impedance migrated to radio and then to recording equipment. Other impedances like the 75 ? coax for video and digital audio is unbalanced which causes more ground noise pick up.
The transfer mechanism is leakage currents that flow on the signal cables, which is a shield if the connections are unbalanced. The coupled noise is due to Ohm’s law, the induced voltage is equal to the current times the resistance, and a shield is a much higher resistance than a wire conductor. On essentially, only the shield resistance matters. All unbalanced interfaces have ground coupling issues so unbalanced interfaces and professional equipment cannot coexist.
A balanced line is the best solution for noise protection. Balance does not mean symmetry. In fact, if the load impedance is 10 times the source impedance, then 90 percent of the signal is transferred. Balanced impedances to ground, on the other hand, minimize noise pick up because most of the noise is common mode and therefore easily rejected by a differential input. In addition, twisting the signal pairs reduces EMI coupling.
Analyzing the interface as a Wheatstone bridge illustrates the noise coupling issue. If the bridge resistors are equal, this represents the worse case situation, since the smallest differential creates a large voltage across the span. By using a low source and high receiver impedances the signal coupling is maximized and the noise coupling is minimized. One way to get this impedance ratio is to use a transformer since its input impedance is in the 100’s of k ??and has an inherently high common mode rejection ratio (CMRR).
Equipment designers need to consider noise rejection throughout the entire interface. The total dynamic range of the interface is equal to the SNR + additional headroom. If an interface has a CMRR of 80dB then the TDR would be in the range of 100 dB, about the same as a CD. A device with an input transformer can easily get more than 100 dB CMRR.
For active input stages, most have impedances of about 20 K ohms. Unfortunately, a low noise interface needs something greater than 200 M ohms. There are ways to configure differential amplifiers to get the high impedances, but don’t look into the TI app notes for good advice on differential amplifiers. The problem is that most differential amplifier designs try to get a high input impedance without considering the input offset current. A typical instrumentation amplifier design has to use fairly small input resistors to manage the gain parameters. To solve this problem, the use of a feedback amplifier to bootstrap the input stage raises the input impedance and eliminates the offset current.
RFI protection is another area where the interfaces matter. By bootstrapping the input capacitances, the line drivers can bypass the rf interferences. There are devices on the market that are universal balanced output drivers that can minimize induced and conducted noise over the signal lines. Shielding and filtering through low pass filters at all I/O ports helps to protect against RFI.
Along with the drivers, designers have to consider cable shielding. It is best to connect the shields at only one end, preferably at the driver side. This practice helps to reduce shield current induced noise, as there is no current path. However, the problem is that pin 1 of an XLR is defined as signal reference ground in the equipment. Designers can fix this problem by making a separate path to ground inside the equipment. A hybrid shield ground allows the input to be shielded via a capacitor connection to ground.
While implementing grounding systems, one must consider the power supply. Tie the power supply to the chassis on the shortest path possible to allow for zero magnetic loops and minimal resistance to ground. Magnetic loops affect equipment through flux density. It’s important to reduce any loop areas in external locations and keep signal lines as far away as possible from any magnetic components.
